CheXpert-5-convnextv2-tiny-384

This model is a fine-tuned version of facebook/convnextv2-tiny-22k-384 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.1644
• Auroc Atelectasis: 0.4756
• Auroc Cardiomegaly: 0.6956
• Auroc Consolidation: 0.5969
• Auroc Edema: 0.8883
• Auroc Pleural effusion: 0.8692
• Specificity Atelectasis: 0.0816
• Specificity Cardiomegaly: 0.7215
• Specificity Consolidation: 0.7068
• Specificity Edema: 0.4835
• Specificity Pleural effusion: 0.2532
• Exact Match: 0.0402
• Hamming Distance: 0.4554

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.001
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 2500
• num_epochs: 6

Training results

Training Loss	Epoch	Step	Validation Loss	Auroc Atelectasis	Auroc Cardiomegaly	Auroc Consolidation	Auroc Edema	Auroc Pleural effusion	Specificity Atelectasis	Specificity Cardiomegaly	Specificity Consolidation	Specificity Edema	Specificity Pleural effusion	Exact Match	Hamming Distance
0.1201	1.0	2010	0.1180	0.7583	0.8187	0.6532	0.8171	0.8295	0.1447	0.4273	0.9971	0.3333	0.0811	0.1346	0.3061
0.1177	2.0	4020	0.1176	0.7586	0.8387	0.6655	0.8196	0.8411	0.3666	0.6803	0.7962	0.3031	0.2830	0.1541	0.2769
0.1145	3.0	6030	0.1124	0.7736	0.8423	0.6789	0.8345	0.8541	0.4841	0.6914	0.9640	0.2092	0.2477	0.2013	0.2523
0.1098	4.0	8040	0.1094	0.7852	0.8619	0.6923	0.8407	0.8629	0.3826	0.8147	0.7786	0.3836	0.1550	0.1719	0.2590
0.104	5.0	10050	0.1062	0.7904	0.8632	0.7045	0.8515	0.8736	0.4176	0.7128	0.8683	0.3822	0.2096	0.1955	0.2496
0.0984	6.0	12060	0.1061	0.7944	0.8652	0.7057	0.8536	0.8741	0.4495	0.7491	0.8963	0.4333	0.2879	0.2398	0.2349

Framework versions

• Transformers 4.48.3
• Pytorch 2.1.2
• Datasets 3.3.0
• Tokenizers 0.21.0